A group multi-criteria decision-making preference learning method considering conflicting opinions is proposed for multi-criteria group decision-making problems in a probabilistic linguistic environment. First, based on distance measures, the differences in expert evaluations are measured, and expert weights are determined according to the consistency between individual and group evaluation matrices. Then, to address the problem that the traditional Utility Additive (UTA) method has no solution in group decision-making due to the conflict of experts" preferences, a two-stage conflict resolution method based on expert weights and maximum consistency constraints is proposed. In the first stage, according to expert weights, the candidate preference sets are identified using the weighted majority principle; in the second stage, a 0-1 integer programming model with the goal of maximizing the retention of the preferences of experts with high weights is constructed to screen out the largest compatible consistent preference subset. Furthermore, the UTA model is solved according to the maximum consistency preference subset to obtain the marginal utility function, and the ranking results of each alternative are obtained based on the marginal utility function. Finally, taking the problem of the location selection of pension institutions in a certain city as an example, the feasibility and applicability of the proposed method are verified.